Re: Possible bug in zfs send or pipe implementation?

Confirmed. Removing PV from the pipeline has solved my issue. I was able to move 20TB around in the last 18 hours without issue.

Sent from my iPhone

> On 15 Jul 2024, at 6:42 PM, Jason Tubnor <jason@tubnor.net> wrote:
> 
> 
> 
> 
> On 15/07/2024 9:35 am, Garrett Wollman wrote:
>> I'm currently running syncoid with `--quiet`, which removes `pv` from
>> the pipeline.  It seems to be moving along, but of course I can't
>> easily tell what it's doing.
>> 
>> Without `pv` in the way, the process reading the pipe is `lzop`
>> instead, which doesn't try to do any fancy select() stuff, it's just a
>> normal filter reading data from stdin and writing compressed data to
>> stdout, in sequence.
> I'm having the same issue here. I thought it was just my hardware and/or syncoid that was having this issue on 3.5TB dataset replications.
> 
> I'll try the quiet mode to see if removing PV solves the problem. I'm also using compression=none so I won't see lzop.
> 
> Cheers.
> 
>>